Fourteen modules of theory, mechanics, platforms, psychology, and case studies—and most preventable losses still come from the same dozen mistakes, repeated with new tickers each cycle. This chapter closes Case Studies & Post-Mortems with a mistake catalog in prose: symptom, mechanism, composite example, and fix. Use it when hindsight whispers that you “always knew.”
No new math here; only enforcement of sizing, signals, calibration, and bias systems you already read.
The twelve mistakes (overview)
Trading vague resolution yields surprise INVALID outcomes. Confusing display mid with tradable probability buys automated-market-maker lies. Ignoring fees turns paper edge into live losses. Full Kelly without caps invites ruin streaks. Correlated trees sized as independent coins wipe twenty-five percent nights. Market orders on spikes pay debate slippage. Chasing cross-venue gaps without rule diffs loses the “free” six cents. Updating forecast after seeing price anchors to the screen. Endowment keeps dead positions because “it’ll come back.” Recency trades headlines at T0. Overconfidence sizes ninety percent on eighteen-cent long tails. No journal or Brier means repeating the same leak forever.
The case studies from 2020 through 2024 are illustrations of these twelve wearing different headlines.
Resolution surprise
A trader buys YES at sixty-two cents on “Candidate wins election” without specifying office or certifier path. The market resolves INVALID with partial refund—roughly forty cents per share versus expected binary loss. Mechanism: ambiguous specification, not rigging. Augur-era disputes echo: rules beat conviction.
Fix: copy resolution text into the journal; list three failure modes; if any subjective risk exceeds five percent, halve size or pass.
AMM mid as gospel
Display shows forty-two percent on a viral prop; three thousand dollars market-buys effectively near fifty-one; media quotes forty-two. Mechanism: curve convexity plus possible pump; conditional efficiency fails.
Fix: walk the simulator; if slippage exceeds three cents versus display, do not cite and do not size as if forty-two were tradable.
Debate-night market order
Gaffe headline; price forty-eight to fifty-seven in four minutes; trader buys fifty-six, sells fifty-two within the hour. Mechanism: volatility spike plus recency; view may be fine, execution is not.
Fix: cooling period; limit only at pre-spike thesis price.
Correlation cluster blow-up
Senate, president, and three swing props all YES; one uniform swing loses twenty-two percent of bankroll. Mechanism: one tree treated as five coins.
Fix: twenty-five percent open-risk cluster cap; haircut table in portfolio case study.
Cross-venue arb without rules
Kalshi YES fifty-four, offshore forty-eight; trader buys offshore, sells Kalshi; offshore resolves NO on technicality while Kalshi pays YES. Mechanism: different events, same headline.
Fix: side-by-side resolution diff; if not byte-match, no arb.
Mistakes and bias overlap
Spike chasing pairs recency with affect—cooling rooms address both. Anchoring after price pairs with confirmation—blind initial forecasts address both. Endowment pairs with loss aversion—re-buy audits address both. Overconfidence on niche legs pairs with affect—probability checks and shave maps address both. No journal enables hindsight—immutable logs and Brier address it.
After a losing week
List every trade with tags. Match top two to the twelve mistakes. Count gate violations. Recompute Brier on locked forecasts, not P&L stories. Install one friction—limit-only week, for example. Shadow the next five ideas; log at least three.
If the same top mistake repeats two weeks, halve unit size until ninety percent gate compliance returns.
Sixty-second pre-flight
Does resolution PDF match this venue? Is edge net of fees above threshold? Is size within Kelly cap times correlation haircut? Is the order a limit with a written thesis stop? Is emotion score three or below on a five-point scale? Fail any gate, pass.
Same edge, two traders
Systems trader: fifty-seven percent blind scout forecast, three cents net edge, 1.4 percent bankroll, limit fifty-five, no action on debate spike, month plus four hundred twenty. Mistakes trader: fifty-seven percent after seeing fifty-four, “six cents” gross edge, eight percent conviction, market fifty-four, buys fifty-nine on spike, month minus nineteen hundred.
Same view, different process grade—Module 14 is about process.
When builders and media err
Journalists quote thin mids. Operators list ambiguous props. Protocols lack dispute runbooks. Growth teams incentivize volume over depth. Traders still pass.
Arc from case studies to checklist
2020 taught resolution and caps. Midterms taught trees and waves. Celebrity and ETF cases taught semantics and dockets. Augur taught oracle risk. Portfolio and manipulation cases taught correlation and depth. 2024 taught surge execution. This chapter names the reusable mistakes so the next module—future of prediction markets—starts from habits, not hope.
Mistakes seven through twelve in prose
Chasing a six-cent gap between venues without a byte-matched resolution diff is mistake seven: you thought you arbitraged, you bought two different events. Updating your forecast only after you saw the screen is mistake eight: the market moved you, you did not lead. Holding because you are already in is mistake nine: endowment dressed as patience. Buying the headline thirty seconds after it prints is mistake ten: recency with extra fees. Sizing ninety percent confidence on an eighteen-cent long tail is mistake eleven: overconfidence on thin books. Skipping the journal is mistake twelve: you will tell yourself a story next month that your log would disprove.
Each mistake has a case-study scar in Module 14; this list is the index.
Weekly scorer prompts
What were my top two mistake tags? Did I violate any gate? Which forecast was locked before the trade? Did slippage exceed plan? One friction to add next week—only one, or you will install none.
Right view, wrong process: the module’s theme
Case studies from 2020 through 2024 repeat one plot: two traders share a belief, one uses limits and locked forecasts, one uses market orders and screens. The second trader tells a better story at the bar; the first trader has better ledgers. Module 14 is not a victory lap for markets; it is a victory lap for process when process is enforced.
Builders and journalists: mistakes that hurt traders
Quoting thin mids is mistake thirteen for media. Listing ambiguous props is mistake fourteen for operators. Traders cannot fix those, but you can refuse to trade broken products and refuse to share undepth screenshots that train bad habits in readers you influence.
What to carry forward
Twelve mistakes explain most hobbyist and semi-pro leaks—resolution and execution top the list. Composite examples show right view, wrong process is common. Weekly post-mortem: tag, mistake number, one friction, shadows. Turn prior modules into defaults at the gate, not bedtime reading.
Next: Module 15 — The Future of Prediction Markets: EventFi, AI versus crowds, privacy, mainstream adoption, and regulatory paths ahead.